RECOGNITION SYSTEM USING MYO ARMBAND FOR HAND GESTURES - SURVEY

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 10 Issue: 05 | May 2023 www.irjet.net p-ISSN: 2395-0072
© 2023, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 1736
RECOGNITION SYSTEM USING MYO ARMBAND FOR HAND GESTURES -
SURVEY
PAVAN K S 1, Prof. TEJASWINI. P. S 2
1 M. Tech student, CSE, Bangalore Institute of Technology, Karnataka, India.
2 Professor, Dept. of CSE, Bangalore Institute of Technology, Karnataka, India.
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract -In previous years, therecognitionsystemforhand
gestures has been researched greatly as it has the ability to
communicate and connect with any IoT device fruitfully with
the help of the Human-computerinterface. Thispaperpresents
an overview of the recent recognition systems using Myo
armbands for hand gestures. Common methods used in the
recognition system for hand gestures of the referred papers
are listed in this paper. A summary of the recognition system
for hand gestures with methods, databases, and algorithms
used with the aim of the paper and some drawbacks and
future work are mentioned.
KEYWORDS: U-Net, CNN, ANN, RNN, Deep Learning, Myo
Armband.
1.INTRODUCTION
The recognition system for hand gestures aim is to
implement a natural interaction between humans and
computers in a way in which the registered gestures are
used in controlling the robotic system or transferring useful
information. This research is useful for people who are
suffering from hearing loss and people who can gain a
helpful way to use input gadgets to obtain peak-level of
accuracy while maxing out the convenience at which the
users make use of it. As they rely on hand gestures for
communication they have to be able to convert any gesture-
based language into a text form with the help of recognition
gadgets that are used for translations. [1] TheMyoArmband
for controlling the gesture is one of the few devices used in
identifying Hand gestures that usually represents the user’s
intention to act upon or achieve specific behaviors using
devices equipped to the arm to easily capture the data and
interpret them. [2]
This device is a new approach to the process of human-
computer interaction (HCI) tool that is implemented to
capture the forearm’s muscular activity of the person its
equipped to and then interprets the gestures of the fingers,
hands, and arms by analyzing EMG (electro-myo-gram)
signals of the muscles.[3] In this, they use EMG sensors for
the effective capture of EMG signals.[2][3]
According to the data collected by World - Health -
Organization, the values of people suffering due to hearing-
loss is approximately more than 1.5 billion people (nearly
20% of the human population in theworld)livewithhearing
loss in 2022, 430 million of them have disablinghearingloss.
It is expected that by 2050, people diagnosed with hearing
loss may cross 180% of the current cases.[5] The number
was 278 million in 2005.[4] Hence, the requirement of a
recognition system for hand gestures that the hearing
impaired may use to easily express their thoughts, feeling,
and knowledge instead of verbal communication is
important. In these studies, it has been shown that it is not
efficient to construct a basic EMG-based recognition system
for hand gestures as the problem is complex because of the
various gestures being used, and EMG systems having
multiple electrodes.[8] Thus, to build an efficient solution,
algorithms/methods of machine learning or deep learning
are used to interpret the underlying pattern of the signals to
recognize hand gesturesautomatically.Toaccomplishsucha
goal, k-NearestNeighbor(kNN)[8],Decision-Treealgorithm,
Support-Vector-Machine algorithm (SVM) [6][8], and
Artificial-Neural-Networks algorithm (ANN) [8] are
employed as they are effective in this process. This method
can even be of use to the movement of the prosthetic hand
with the attachment of the Myo armband to the arm that
helps control gestures of the dexterous prosthetic hand by
using arm muscles data.[7][13] The data used is tested by a
10-fold cross-validation method to compare the learning
algorithms with each other(“which are as follows Random
Forest1, RBF-Network2, Neural-Network3, Naïve-Bayes4,
KNN5, NB-Tree6, J487, and Decision Table8 that can be some
of the algorithms”). This is done to obtain the data for the
Myo armbands to be trained to recognize.[9] Recent studies
have explored supervised learning-based methods, such as
CNN (convolutional neural network) and RNN (recurrent
neural network) to implement the HCI (Human-Computer
Interaction) device.[10] We can make use of deep-learning
techniques for feature extraction to be done automatically
and classification isalsodonethesameway.[12]Thismethod
of recognition is utilized in a variety of topics that includes
the recognition of sign language, prosthesis control of the
human limb or arm, controlling the robots, and interfaces
used in games/VR.[13]Currently, in 2020, using
Photoplethysmography to recognize the data of hand
gestures.[14] Then U-net was used for hand image
processing to recognize the gestures more accurately with
respect to hand movements.[15] The UNET architecturecan
be used to access the semantically segmented mask of the
input, which is then given to a VGG16 model for
classification.[16]

2.COMMON STEPS IN RECOGNITION SYSTEM FOR
HAND GESTURES:
2.1.Preprocessing the data
This method starts with the identification and
analysis of signals by calculating the absolute value in its
channels by using every value [12]. This process is done to
cancel out any noise present in thesampleobtained.Inorder
to gain the proper data to train we use certain methods to
filter the noise in the image or any data obtained.
2.2.Feature Extraction
A feature is an attribute of the observed process. Using
this process, the above mentioned algorithms can easily
perform classification [14]. The best way of feature
extraction can be done with a proper segmentation method
which can lead to successful recognition.
2.3.Classification
After using the method of feature extraction, a certain
group or set of feature(s) may degrade or may add no value
to the classifier performance, then thismayleadtoobtaining
a good measure for feature selection by classifying the
number of times a feature splits a tree.[14]The algorithms
used are described in the summary table section of this
paper. As compared, the best classifieralgorithmisthedeep-
learning methods and eventheSVMmethod whichyield high
accuracy.
2.4.Data Collection:
The data set can be obtained from some organization or
from a group of individuals who can gain the required data
for the process needed. In the table of this paper, we have
listed the survey details that are used by different
researchers.
2.5.Recognition:
After the analysis of hand images used as input, the
classification of gestures is used to recognize gestures. This
process is done with the proper selection of attributes or
parameters and a suitable algorithm for classification. Here,
the analysis of the recognition system is done to check the
accuracy of the model [12].
3.AREAS WHERE WE USE THIS SYSTEM
This system was applied for various applications in
multiple areas of expertise, as mentioned in [1][4][6][7][10]
including; translation of hand signal language, VR-
environment, control of robots, in the medical field, etc.
3.1.Recognition of Hand Signal Language: Since hand
signal language is used for communication, it is effectively
researched [1]. There are numerous systems to recognize
gestures for various types of signal languages that are used
[4]. For example [1][4] ASL to be recognized, use a model
with that feature.
3.2.Control of Robots: Being able to control the machine
with the help of gestures is one of the best applications of
this system.[6][10].
3.3.Recognition of numeric system: Recognizing numbers
with the hand gestureisanotherapplication.[8]For example,
the way to recognize the number used in the signal
languages.[8]
4.MATERIALS REQUIRED
4.1.Myo Armband
This is the device that the recent papers used in the
recognizing systems used for handgesturestoberecognized
with EMG. The figure below shows Myo Armband, is a digi-
device produced by Thalmic Labs. It has 8 EMG sensors.This
component has an Inertial-Measurement-Unit (IMU).
Fig.1. “Myo-armband”( “by Thalmic Labs”)
5.SUMMARY TABLE
PAPER Algorithms
Used
Test Data Set
-----------------
Database used
Level of
Accuracy
Obtained
[1]
1.MAST
(combination of
MAST and
Dynamic
random forest)
3000 gestures for
testing
--------------------
American Sign
Language(ASL)
89% -
93%
[2] 1.cross-
correlation
coefficients
(ACCC)
2.k-nearest
neighbor (K-
NN)
3. support-
vector-
machines
(SVMs)
Own dataset
obtained through
6 volunteers
Divided into
24.866 samples
79% -
81%

[3]
1. Artificial
neural
network(ANN)
2. Kohonen
SOFM (self-
organizing
feature map)
3. adaptive
fuzzy k-NN
classifier
(AFMC)
4.SVM
5. Recurrent
Neural Networks
(RNN)
1106 test data
60% -
90%
76% -
83%
96.6%
70%-
100%
[4] 1.Support -
Vector-
Machine (SVM)
2. Ensemble
Learning
(Bagged Tree)
5200 gestures
Of which
4050 used for
testing
-------------------
-
Based on
American Sign
Language(ASL)
own database
60.85%
80%
[6] 1. Support -
Vector -Machine
(SVM)
Size of data 400 84% - 92%
[7] 1. Support -
Vector -Machine
(SVM)
50 gestures for
testing
With 9
repetitions
50% - 97%
[8] 1.k-Nearest
Neighbour
2.Support
Vector Machine
(SVM)
3.Artificial
Neural
Networks(ANN)
Number data
set from 0-9 in
TSL database
i.e., 10
gestures
and 55000
data points
87%
[9] 1.Naïve Bayes,
2.Neural
Network,
3.RBF
Network,
4.K-Nearest
Neighbor,
5.NB Tree
6. J48,
7.Decision
Table,
8.Random
Forest.
562 samples 61%
52% - 53%
62% - 64%
52%
53% - 55%
51% - 53%
53%
55% - 59%
[10] 1.convolutional
neural network
(CNN) model
with deep Q-
Network(DQN)
2.long short
term memory
(LSTM) model
with DQN
3 gestures,
Each has 30
training data
and 20 test
data
Total 90
training data
and 60 test
data
98.33%
[11] 1.Kalman filter
algorithm
2.the neural
network
classification
30,361 data set
from different
operator’s
hand activities.
60% is used as
training data
randomly.
99%
[12] “Deep-learning
techniques”
1.convolutional
neural network
(CNN)
2.Artificial
Neural
Networks(ANN)
5 gestures,
37200 samples
data set,
7500 samples
were used for
testing.
85.08% ±
15.21%
[13] 1.Artificial
Neural
Network
(ANN),
2.Support
Vector Machine
(SVM),
3.k-Nearest
Neighbours (k-
NN), 4.Linear-
Discriminant-
Analysis (LDA)
5.Random
Forest
5 gestures,
7800 total
data,
200 in a
sample
Test data 39
used for each
gestures.
94.8%
[14] 1.Support
Vector Machine
(SVM)
4 gestures,
775 data
samples,
77% for
training and
rest is for
testing.
92% - 95%
[15] 1.deep learning
models
2. “Deep-neural
networks” like
U-Net,
Seg-net,
FCN.
Two different
datasets:
1.Egohands
15000 images
2.GTEA
4800 images
are used.
70% data for
training, 20%-
validation and
10% -testing.
98% -
Egohands,
and
90%- GTEA

[16] 1.CNN-based
2.FCN
3.U-Net
Brazilian Sign
Language
database, with
9600 images
98.97%
6.DRAWBACKS AND FUTURE WORK
As the systems are in the process of being improved further,
they are usually expensive and complex [7]. Another
drawback to check for is the recognition done by a single
hand and check for overlapping images and blurred images.
The EMG signals vary from one to another; it changes from
one particular time to another in the same person [12].
Dynamic gestures can be used fordifferentorientations[13].
There can be further potential advantages of adding an
accelerometer to the developed band [14]. The drawback of
U-Net is the time taken to train it is huge. So for larger
images, more GPU - memory is needed [15].
7. CONCLUSIONS
The process of data obtaining and the process of
selection of algorithm which is for gesture recognition may
rely on the research being done and the application needed
to establish that as one of the working model. In this survey
the areas of recognition system used for hand gesture which
is presented in papers are discussed. Explanation of
recognition system drawbacks, detail survey on recent
recognition systems using Myo armband for hand gestures.
And deep learning algorithms, U-net based recognition
system for hand gestures are presented. Summary of
recognition systems for hand gestures are listed in a tabular
for with results, methods used, algorithms used and theaim.
In most cases as the summary the commonly used
algorithms like SVM, U-Net, Seg-net, and FCN yield most
accurate result of recognition. Thus, the recognition system
for hand gestures providesaninterestinginteractiondomain
in various computer applications.
REFERENCES
[1] Zuhaib MuhammadShakeel,Soonhyuk So,Patrick Lingga,
and Jaehoon (Paul) Jeong. “MAST: Myo Armband Sign-
Language Translator for Human Hand Activity
Classification”, in IEEE, International conference on
Information and Communication Technology Convergence
2020(ICTC 2020) pp.494-499, September-2020.
[2] Gonzalo Pomboza-Junez, Juan Holgado Terriza. “Hand
Gesture Recognition based on sEMG signals using Support
Vector Machines”, in 2016 International Conference on
Consumer Electronics-Berlin (ICCE-Berlin), pp.177-181,
June-2016.
[3] Amina Ben Haj Amor, Oussama El Ghoul, Mohamed
Jemni. “Toward sign language handshapes recognitionusing
Myo armband”, IEEE 2017/ Research Laboratory LaTICE,
University of Tunis, Tunisia, October-2017.
[4] Celal Savur, Ferat Sahin. “American Sign Language
Recognition System byUsing SurfaceEMGSignal”,2016IEEE
International Conference on Systems, Man, and Cybernetics
(SMC 2016) pp. 2872-2877, October,9-12, 2016.
[5] W. H. O. (WHO), “Deafness and hearing loss, Report,”
World Heath Organization Media Center Fact Sheet, 2022.
[Online]. Available: https://www.who.int/health-
topics/hearing-loss#tab=tab_1 , Date retrieved: 08.05.2023
[6] Karthik Sivarama Krishnan, Akash Saha, Srinath
Ramachandran, Shitij Kumar. “Recognition of Human Arm
Gestures Using Myo Armband for the GameofHand Cricket”,
2017 IEEE International Symposium on Robotics and
Intelligent Sensors (IRIS2017), pp.389-394, 5-7 October.
2017, Ottawa, Canada.
[7] M. Cognolato, M. Atzori, D. Faccio, C. Tiengo, F. Bassetto,
R. Gassert and H. Muller. “Hand Gesture Classification in
Transradial Amputees Using the Myo Armband Classifier”,
2018 7th IEEE International Conference on Biomedical
Robotics and Biomechatronics (Biorob) Enschede, The
Netherlands, pp.156-161, August 26-29, 2018.
[8] Engin Kaya, Tufan Kumbasar. “HandGesture Recognition
Systems with the Wearable Myo Armband”, 2018 6th IEEE
International Conference on Control Engineering &
Information Technology (CEIT), 25-27 October 2018,
Istanbul, Turkey.
[9] Tanasanee Phienthrakul. “Armband GestureRecognition
on Electromyography Signal for Virtual Control”, 2018 4th
IEEE International Conference on Business Industrial
Research (ICBIR2018), pp.149-153, September-2018,
NakornPahom, Thailand.
[10] W. Seok, Y. Kim, C. Park. “Pattern Recognition of Human
Arm Movement Using Deep Reinforcement Learning”, 2018
IEEE International Conference on Information Networking
(ICOIN 2018), pp.917-919, Feb-2018.
[11] Jingxiang Chen, Chao Liu, Rongxin Cui,ChenguangYang.
“Hand Tracking Accuracy Enhancement by Data Fusion
Using Leap Motion and Myo Armband”, 2019 International
Conference on Unmanned Systems andArtificial Intelligence
(ICUSAI 2019), pp.256-261, April-2019.
[12] Edison A. Chung, Marco E. Benalcázar. “Real-TimeHand
Gesture Recognition Model UsingDeepLearningTechniques
and EMG Signals”, IEEE 2019 27th European Signal
Processing Conference (EUSIPCO 2019), September-2019.
[13] Sam Young, Benjamin Stephens-Fripp, Andrew Gillett,
Hao Zhou and Gursel Alici. “Pattern Recognition for
Prosthetic Hand User’s Intentions using EMG Data and

Machine Learning Techniques”, 2019 IEEE/ASME
International Conference on Advanced Intelligent
Mechatronics Hong Kong, China, pp.544-550, July 8-12,
2019.
[14] Karthik Subramanian, Celal Savur, Ferat Sahin. “Using
Photoplethysmography for Simple Hand Gesture
Recognition”, IEEE 15th International Conference of System
of Systems Engineering (SoSE 2020), pp.307-312, June 2-4,
2020 Budapest, Hungary.
[15] Akhila Sreekumar, Geetha M. “Hand Segmentation In
Complex Background Using UNet”, 2020 IEEE 2nd
International Conference on Inventive Research in
Computing Applications (ICIRCA-2020), pp.440-445, Feb-
2020.
[16] H Pallab Jyoti Dutta, Debajit Sarma, M.K. Bhuyan and R.
H. Laskar. “Semantic Segmentation based Hand Gesture
Recognition using Deep Neural Networks”,2020 IEEE
International Conference on Research and Education in
Mechatronics (REM 2020), May-2020.

RECOGNITION SYSTEM USING MYO ARMBAND FOR HAND GESTURES - SURVEY

Recommended

Recommended

More Related Content

Similar to RECOGNITION SYSTEM USING MYO ARMBAND FOR HAND GESTURES - SURVEY

Similar to RECOGNITION SYSTEM USING MYO ARMBAND FOR HAND GESTURES - SURVEY (20)

More from IRJET Journal

More from IRJET Journal (20)

Recently uploaded

Recently uploaded (20)

RECOGNITION SYSTEM USING MYO ARMBAND FOR HAND GESTURES - SURVEY